Skip to content
No PriorsNo Priors

No Priors Ep. 97 | With Decagon CEO and Co-Founder Jesse Zhang

Today on No Priors, co-founder and CEO of Decagon, Jesse Zhang, joins Elad to discuss the future of agentic customer support. Decagon provides AI-powered customer interactions for companies like Rippling, Notion, Duolingo, Classpass, Substack, Vanta, Eventbrite, and more. Jesse shares the thesis behind starting Decagon, why he sees customer support as the ideal entry point for agentic technology, and what areas of AI excite him most. They also discuss voice-based interfaces, issues with latency in current capabilities, and the connection between young math olympiad communities and today’s AI startups. Sign up for new podcasts every week. Email feedback to show@no-priors.com Follow us on Twitter: @NoPriorsPod | @Saranormous | @EladGil | @TheJesseZhang Show Notes: 0:00 Introduction 0:30 Starting Decagon 3:15 Business impact of adopting agents for customer support and customer ops 8:00 AI infrastructure and models for customer success agents 12:05 Voice-based capabilities and text-to-speech engines 15:00 Combatting latency 16:25 Crossover of math and AI communities 21:12 Exciting areas of AI 25:29 Strengths and weaknesses of agents

Elad GilhostJesse Zhangguest
Jan 15, 202530mWatch on YouTube ↗

At a glance

WHAT IT’S REALLY ABOUT

Decagon’s Transparent AI Agents Redefine Enterprise Customer Support at Scale

  1. Decagon CEO and co-founder Jesse Zhang discusses building enterprise-grade generative AI agents focused on customer support, already deployed at companies like BILT Rewards, Rippling, Notion, and Duolingo.
  2. He explains that Decagon’s edge lies less in owning core LLMs and more in orchestration, transparency, and software around the models—giving enterprises control, observability, and clear ROI.
  3. Zhang shares concrete impact metrics, such as BILT Rewards saving the equivalent of 65 support agents while improving customer experience and response speed across channels including emerging voice agents.
  4. He also outlines where AI agents will win first—use cases with quantifiable ROI and safe, incremental rollout—and where adoption will be slower due to risk, trust, and measurement challenges.

IDEAS WORTH REMEMBERING

5 ideas

Focus on use cases with clearly measurable ROI and incremental rollout.

Decagon chose customer support because you can quantify automation (deflection rates, headcount saved, CSAT/NPS) and safely start with a small traffic slice before scaling.

Transparency and control are critical for enterprise AI adoption.

Large customers demand visibility into what data the agent uses, how decisions are made, and the ability to inspect, audit, and adjust behavior rather than treat AI as a black box.

Most differentiation sits above the base models in orchestration and software.

Since everyone can access similar LLMs, value comes from how you orchestrate multiple models, encode business logic, evaluate performance, and build surrounding tooling and analytics.

Instruction following matters more than pure reasoning in many applied agents.

For customer support workflows, strict adherence to policies and SOPs is more impactful than improved quantitative reasoning, so advances in instruction-following will unlock more automation.

Voice agents are becoming viable but hinge on latency and UX design.

High-quality TTS/ASR and voice-to-voice models from vendors like OpenAI and ElevenLabs are enabling phone-based agents, but latency, streaming strategies, and conversational pacing remain key challenges.

WORDS WORTH SAVING

5 quotes

We kind of arrived at our current use case as maybe what we think is the golden use case for these AI agents, which is customer interactions, customer service.

— Jesse Zhang

Most applications nowadays are real software companies and AI models are kind of tools that everyone can use.

— Jesse Zhang

The thing that’s made us special so far is we have a huge sort of focus on transparency… it’s very important for them that the AI agent is not a black box.

— Jesse Zhang

So far, it’s around 65 agents of just headcount saved… the customer experience is also a lot snappier.

— Jesse Zhang, on BILT Rewards

For the vast majority of use cases right now, there’s not going to be real commercial adoption with the state of the current models.

— Jesse Zhang

Decagon’s product focus: AI agents for customer service and customer experienceEnterprise needs for transparency, observability, and control in AI systemsTechnical architecture: orchestration layers, tooling, and multi-model useVoice-based AI support, latency challenges, and multimodal interfacesMath Olympiad and contest communities as a talent and founder pipelineImpact on customer support organizations: automation, restructuring, and new rolesCriteria for successful AI agent use cases and realistic near-term adoption

High quality AI-generated summary created from speaker-labeled transcript.

Get more out of YouTube videos.

High quality summaries for YouTube videos. Accurate transcripts to search & find moments. Powered by ChatGPT & Claude AI.

Add to Chrome